U.S. patent application number 11/671370 was filed with the patent office on 2007-08-23 for information processing apparatus, method, and program product.
This patent application is currently assigned to Sony Corporation. Invention is credited to SHUNJI YOSHIMURA.
Application Number | 20070198508 11/671370 |
Document ID | / |
Family ID | 38429582 |
Filed Date | 2007-08-23 |
United States Patent
Application |
20070198508 |
Kind Code |
A1 |
YOSHIMURA; SHUNJI |
August 23, 2007 |
INFORMATION PROCESSING APPARATUS, METHOD, AND PROGRAM PRODUCT
Abstract
An information processing apparatus includes an extraction unit
configured to extract sequentially in time, keywords from multiple
character strings, and a similarity calculation unit configured to
calculate similarity values of the extracted keywords included in
adjacent first regions of the multiple character strings, each
first region including a predefined part of the multiple character
strings.
Inventors: |
YOSHIMURA; SHUNJI;
(Shinagawa-ku, JP) |
Correspondence
Address: |
OBLON, SPIVAK, MCCLELLAND, MAIER & NEUSTADT, P.C.
1940 DUKE STREET
ALEXANDRIA
VA
22314
US
|
Assignee: |
Sony Corporation
Minato-ku
JP
|
Family ID: |
38429582 |
Appl. No.: |
11/671370 |
Filed: |
February 5, 2007 |
Current U.S.
Class: |
1/1 ;
707/999.005; 707/E17.028 |
Current CPC
Class: |
G06F 16/7844 20190101;
Y10S 707/914 20130101 |
Class at
Publication: |
707/005 |
International
Class: |
G06F 17/30 20060101
G06F017/30 |
Foreign Application Data
Date |
Code |
Application Number |
Feb 8, 2006 |
JP |
2006-030482 |
Claims
1. An information processing apparatus, comprising: an extraction
unit configured to extract sequentially in time keywords from
multiple character strings; and a similarity calculation unit
configured to calculate similarity values of the extracted keywords
included in adjacent first regions of the multiple character
strings, each first region including a predefined part of the
multiple character strings.
2. The information processing apparatus according to claim 1,
further comprising: a detection unit configured to detect a
boundary point between the adjacent first regions based on a
similarity value of the included keywords being lower than a
threshold similarity value, the boundary point corresponding to
changing topics in a content including the multiple character
strings.
3. The information processing apparatus according to claim 1,
wherein the multiple character strings are configured to be
displayed on a screen by a user and the multiple character strings
correspond to pictures.
4. The information processing apparatus according to claim 1,
wherein the character strings are displayed on a screen when a
content is reproduced.
5. The information processing apparatus according to claim 1,
wherein the similar values are minimal values.
6. The information processing apparatus according to claim 5,
wherein the minimal values are lower than a predefined similarity
threshold value.
7. The information processing apparatus according to claim 1,
further comprising: a detection unit configured to detect a part of
a content in which a similarity value calculated by the similarity
calculation unit is higher than a prescribed similarity value.
8. The information processing apparatus according to claim 1,
wherein the predefined part of the multiple character strings
includes a predefined number of keywords.
9. The information processing apparatus according to claim 5,
wherein the predefined number is 10.
10. The information processing apparatus according to claim 1,
wherein the predefined part of the multiple character strings
includes a predefined number of sentences.
11. The information processing apparatus according to claim 1,
wherein the predefined part of the multiple character strings has a
predefined time span.
12. The information processing apparatus according to claim 1,
wherein a similarity value of keywords is calculated based on an
Euclidian distance or a cosine measure in a vector space model.
13. The information processing apparatus according to claim 1,
wherein a similarity value of keywords is calculated based on a
predefined number of identical keywords identified in the adjacent
first regions.
14. The information processing apparatus according to claim 1,
wherein the similarity calculation unit determines adjacent second
regions for calculating similarity values, each second region
including at least one other keyword than a corresponding first
region.
15. The information processing apparatus according to claim 2,
further comprising: an attribute information generation unit
configured to generate information to be associated with the
boundary point.
16. The information processing apparatus according to claim 15,
wherein the information associated with the boundary point is one
of time information, title information, and topic information.
17. An information processing method, comprising: extracting
sequentially in time keywords from multiple character strings; and
calculating similarity values of keywords included in adjacent
regions of the multiple character strings, each region including a
predefined part of the multiple character string.
18. A computer program product stored on a computer readable media,
the computer program product determining a computer to execute
processing comprising: extracting sequentially in time keywords
from multiple character strings; and calculating similarity values
of the extracted keywords included in adjacent regions of the
multiple character strings, each region including a predefined part
of the multiple character strings.
19. An information processing apparatus, comprising: extraction
means for extracting sequentially in time keywords from character
strings; and similarity calculation means for calculating
similarity values of the extracted keywords included in adjacent
regions of the multiple character strings, each region including a
predefined part of the multiple character strings.
Description
CROSS REFERENCE TO RELATED APPLICATIONS
[0001] The present disclosure contains subject matter related to
Japanese Patent Application JP 2006-030482, filed in the Japanese
Patent Office on Feb. 8, 2006, the entire contents of which is
incorporated herein by reference.
BACKGROUND OF THE INVENTION
[0002] 1. Field of the Invention
[0003] The present invention relates to an information processing
apparatus, a method, and a program product that calculate
similarities of keywords as criteria for detecting positions of
changing topics in contents.
[0004] 2. Description of the Related Art
[0005] In related arts, various techniques for detecting a change
between topics (i.e., changing points of the topics) in contents
such as television programs have been proposed. At the positions of
the changing points of the topics in the content, information
indicating the changing points is added to be used, for example, as
chapter indicators/descriptors when the content is reproduced.
[0006] In JP-A-11-234611, a technique is disclosed, in which a list
of topic-changing words is previously stored in the device, and
when a word from the head of the caption information is the same
word as the topic-changing word stored in the list, a displaying
position in the caption information is detected as a changing
position of topics.
SUMMARY OF THE INVENTION
[0007] The invention has been made in view of the above situation,
and similarities of keywords are calculated as one criteria for
detecting positions of changing points of topics in contents.
[0008] An information processing apparatus according to an
embodiment of the present invention includes an extraction unit
configured to extract sequentially in time, keywords from multiple
character strings, and a similarity calculation unit configured to
calculate similarity values of the extracted keywords included in
adjacent first regions of the multiple character strings, each
first region including a predefined part of the multiple character
strings.
[0009] An information processing method and a program-product
according to other embodiments of the present invention includes
extracting sequentially in time, keywords from multiple character
strings, and calculating similarity values of keywords included in
adjacent regions of the multiple character strings, each region
including a predefined part of the multiple character string.
BRIEF DESCRIPTION OF THE DRAWINGS
[0010] FIG. 1 is a block diagram showing an information processing
apparatus according to an embodiment of the invention;
[0011] FIG. 2 is a view showing an example of detection of a
changing point of topics;
[0012] FIG. 3 is a graph showing changing points of topics
confirmed by a human and calculation results of similarities;
[0013] FIG. 4 is a block diagram showing a hardware configuration
example of the information processing apparatus;
[0014] FIG. 5 is a block diagram showing a function configuration
example of the information processing apparatus;
[0015] FIG. 6 is a flowchart explaining changing point detection
processing of the information processing apparatus;
[0016] FIG. 7 is block diagram showing a function configuration
example of an information processing apparatus;
[0017] FIG. 8 is a flowchart explaining attribute information
generation processing of the information processing apparatus of
FIG. 7;
[0018] FIG. 9 is a flowchart explaining attribute information
generation processing of the information processing apparatus of
FIG. 7;
[0019] FIG. 10 is a graph showing an example of calculation results
of similarities; and
[0020] FIG. 11 is a graph showing another example of calculation
results of similarities.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0021] Embodiments of the invention will be described below, and
the correspondence between constituent features of the invention
and embodiments described in the specification and the drawings is
exemplified as follows. If there is an embodiment disclosed in the
specification and the drawings but not identified as the embodiment
corresponding to a constituent feature of the invention, that does
not mean that the embodiment does not correspond to the constituent
feature. Conversely, if an embodiment is disclosed as the
embodiment corresponding to a constituent feature, that does not
mean that the embodiment does not correspond to other than the
constituent feature.
[0022] When the technique disclosed in JP-A-11-234611 is used, it
is necessary that the list of the topic-changing words be
previously prepared and stored. When the topic-changing word is not
in the head of the caption information displayed, although it
corresponds to the position where the topic is changed when the
technique disclosed in JP-A-11-234611, it is difficult to detect
the changing position of the topics. As a result, a reliability of
the detection in the conventional devices is reduced.
[0023] An information processing apparatus (for example, an
information processing apparatus 1 of FIG. 1) according to an
embodiment of the invention includes an extraction unit (for
example, a keyword extraction unit 32 of FIG. 5) for extracting
keywords, sequentially in time, (also called in the specification,
time series of keywords) from character strings corresponding to
displayed pictures. The character strings may be displayed or
spoken by a person. In addition, the information processing
apparatus includes a calculation unit (a similarity calculation
unit 34 in FIG. 5) for calculating similarities of keywords
included in adjacent regions of the character strings. As will be
discussed next, the adjacent regions are defined on a time basis
for example, or as having prescribed ranges relative to the
character strings.
[0024] The information processing apparatus may further include a
detection unit (for example, a changing point detection unit 35)
for detecting a boundary point between the adjacent regions
(prescribed ranges). The detecting is determined based on a
similarity value (called similarity) of keywords included in the
prescribed ranges. When a similarity value is lower than a
threshold similarity value, based on the similarities calculated by
the similarity calculation unit, the changing point (changing
topic) is determined. As will be appreciated by one of ordinary
skill in the art, the threshold similarity value depends on the
method selected for calculating the similarity value, as discussed
next.
[0025] The information processing apparatus further includes a
detection unit (for example, an attribute information generation
unit 112) for detecting a part of a content in which a similarity
calculated by the similarity calculation unit is higher than a
prescribed similarity value. This detected part is indicative of a
vigorous (highly entertaining/desired) part of a subject of the
content as will be discussed later.
[0026] An information processing method and a program-product
according to an embodiment of the invention includes steps of
extracting time series of keywords from the character strings,
which correspond to the displayed pictures, and calculating (for
example, step S4 of FIG. 6) similarities of the keywords included
in the prescribed ranges adjacent to each other on a time base.
[0027] Hereinafter, embodiments of the invention will be explained
with reference to the drawings.
[0028] FIG. 1 is a block diagram showing an information processing
apparatus according to an embodiment of the invention.
[0029] An information processing apparatus 1 is the apparatus in
which contents, such as television programs and movies, are taken
as input, changing points of topics (subjects) are detected, and
changing points information as information indicating the detected
changing points is outputted as an output. The information
processing apparatus 1 is described later in more details.
[0030] Contents to be inputted to the information processing
apparatus 1 include not only video data and audio data but also
text data (multiple character strings) such as closed caption data
used for displaying captions corresponding to the displayed
pictures on a screen when the content is reproduced, and the
changing points of the topics are detected by using the text data
in the information processing apparatus 1.
[0031] As described later, the changing points information
outputted from the information processing apparatus 1 is used for
adding attribute information to the contents. For example, time
information indicating positions of the detected changing points
(information of chapters) and the like are generated as attribute
information to be added to contents.
[0032] FIG. 2 shows an example of detection of a changing point of
two topics. A case where a content to be processed is a news
program is explained.
[0033] In the example of FIG. 2, when the news program is played,
news about "innocent ruling to the defendant Mr. Yamada" is
reproduced as topic 1 from a time "t.sub.1," and news about "new
relief measures for disaster victims have been settled" is
reproduced as topic 2 from a time "t.sub.2". Therefore, a changing
point between topics 1 and 2 of the news corresponds to time
"t.sub.2."
[0034] The content includes closed caption data ("CC" in the
drawing) (text data). When the news about the topic 1, which is
started from the time "t.sub.1" is reproduced, for example,
captions of the same subject as the subject spoken by a caster
about the topic 1 are sequentially displayed on a reproducing
screen, so as to correspond to displayed pictures "P.sub.1" to
"P.sub.m," which represent a subject of the topic 1. The captions
are based on text data in a range from the time "t.sub.1" to
"t.sub.2" in FIG. 2.
[0035] When the news about the topic 2, which is started from the
time "t.sub.2" is reproduced, for example, captions of the same
subject as the subject spoken by a caster about the topic 2 are
sequentially displayed on the reproducing screen, so that the
captions correspond to displayed pictures "P.sub.1" to "P.sub.n,"
which represent a subject of the topic 2. The captions are based on
text data in a range from the time "t.sub.2" to "t.sub.3" in FIG.
2.
[0036] In the case that the content to be processed is similar to
the one described above, time-series of keywords ("KW series" in
the drawing) are extracted from the text data as shown by a tip of
an outline arrow in FIG. 2 in the information processing apparatus
1.
[0037] More specifically, at the start of the news about the topic
1, when the caption "First, this is news about innocent ruling to
the defendant Mr. Yamada." is displayed based on the text data,
time series of keywords "news," "innocent," "ruling," "defendant,"
"Yamada" and so on are extracted from the text data. Similarly, at
the start of the news about topic 2, when the caption "Next, this
is news that new relief measures for disaster victims have been
settled." is displayed based on text data, time series of keywords
"news," "relief measures," "disaster," "victims," and the like are
extracted from the text data.
[0038] When the time series of keywords are extracted, detection
windows respectively surrounding prescribed ranges of keywords,
which are adjacent to each other on a time base, are set in the
information processing apparatus 1. For example, when one detection
window surrounds a range of 10 keywords, if the whole time series
of keywords includes 1 to "N" keywords, a detection window "A,"
which surrounds keywords 1 to 10 and a detection window "a," which
is the detection window surrounding keywords 11 to 20 are set. The
detection window "a" is adjacent to the detection window "A" on the
time base.
[0039] Similarly, a detection window "B," which surrounds keywords
2 to 11, and a detection window "b," which is the detection window
surrounding keywords 12 to 21, are set adjacent to each other on
the time base. A detection window "C," which surrounds keywords 3
to 12, and a detection window "c," which is the detection window
surrounding keywords 13 to 22, are set adjacent to each other on
the time base. That is, the detection windows surrounding keywords
of prescribed ranges are set at adjacent positions by shifting
continuously or not a keyword number by one. Optionally, the
keyboard number may be shifted with a value larger than 1.
[0040] In the information processing apparatus 1, when the
detection windows are set, a similarity value of all keywords
included in a first detection window is calculated by comparing
with an adjacent second detection window on the time base. As
described later, the similarity value may be detected based on the
number of corresponding keywords included in first and second
detection windows. Various techniques for calculating the
similarity value of two detection windows are discussed below.
[0041] In the information processing apparatus 1, the changing
point of two consecutive topics is detected based on the calculated
similarities values as discussed above. For example, a boundary
point (changing point of topics) between adjacent detection windows
is determined when the calculated similarity value is lower than a
preset threshold similarity value.
[0042] In the example shown in FIG. 2, a similarity between
keywords included in a detection window "W.sub.1" and keywords
included in a detection window "W.sub.2," which are set at adjacent
positions on the time base is high, whereas a similarity between
keywords included in a detection window "W.sub.3" and keywords
included in a detection window "W.sub.4" is low. In addition, a
similarity between keywords included in a detection window
"W.sub.5" and keywords included in a detection window "W.sub.6" is
high. The terms "high" and "low" are defined in an exemplary manner
below.
[0043] For example, in one embodiment as shown in FIG. 3, eight
distinct topics are present. A border between two adjacent topics
(turning point) is defined by a minimum value of a cosine measure
of the detected keywords in two adjacent detection windows. In
addition, the calculated minimum value may be required to be
smaller than a predetermined threshold value, for example 0.1 in
FIG. 3. Thus, a minimum value that corresponds to sentence number
10 in FIG. 3 is eliminated as a turning point. Further, the minimum
value may be required to be smaller than a product between a depth
and a positive coefficient k. The depth is defined as a gap between
(1) an average of two maximum values adjacent to a given minimum
value, and (2) the given minimum value. The coefficient k may be
1.0.
[0044] In one embodiment, the changing point is determined by
calculating each minimum value, verifying that the calculated
minimum value is below a threshold value (for example 0.1), and
also verifying that the minimum value is smaller than the depth
multiplied by a coefficient k (for example 1.0).
[0045] In other words, when a detected minimum value satisfies one
or more of the above noted conditions, a similarity value of two
detection windows is considered low. When no minimum value is
detected between two detection windows, the similarity value of the
two detection windows is considered high.
[0046] When the set detection windows are both windows surrounding
keywords extracted from text data concerning the same topic, the
original topic is common to respective windows. Therefore, similar
keywords are included in respective detection windows and, in this
case, the similarity of the keywords included in the detection
windows set at the adjacent positions on the time base is high.
[0047] In the example shown in FIG. 2, the detection window
"W.sub.1" and the detection window "W.sub.2" are both windows
surrounding keywords extracted from the text data concerning the
topic 1. Similarly, the detection window "W.sub.5" and the
detection window "W.sub.6" are both windows surrounding keywords
extracted from the text data concerning the topic 2.
[0048] On the other hand, when the set detection windows are both
windows surrounding keywords extracted from text data concerning
different topics, that is, when the detection window set at a
previous position on the time base is the window surrounding the
keywords extracted from the text data concerning the first topic,
and the detection window set at a position next to the above
detection window on the time base is the window surrounding the
keywords extracted from the text data concerning the second topic,
the original topics are different. Therefore, different keywords
are included in the respective detection windows and, in this case,
the similarity of the keywords included in the detection windows
set at the adjacent positions on the time base is low.
[0049] In the example shown in FIG. 2, the detection window W.sub.3
is the window surrounding the keywords extracted from the text data
concerning the topic 1, and the detection window W.sub.4 is the
window surrounding the keywords extracted from the text data
concerning the topic 2. That is, the detection window W.sub.3 and
the detection window W.sub.4 are windows surrounding keywords
extracted from text data concerning different topics.
[0050] Accordingly, in the information processing apparatus 1,
similarities of respective parts in the content are calculated by
using the text data which is considered to reflect topics of the
content. In addition, the changing point of topics in the content
is detected based on the calculated similarities.
[0051] Accordingly, the changing point of topics can be detected
based on the text data in a more reliable manner as compared with a
case in which the changing point of topics is detected by analyzing
pictures displayed on the screen when the content is
reproduced.
[0052] FIG. 3 is a graph showing changing points of topics visually
checked by a human and the calculated results of similarities found
according to the above method for confirming the effect of a method
according to one embodiment of the invention.
[0053] In the example shown in FIG. 3, a horizontal axis represents
sentence numbers (the text data displayed as captions is delimited
according to sentences and the sentences are sequentially
numbered), and a vertical axis represents similarities (in this
embodiment, the cosine measure based on the later-described vector
space model is used). Values on the vertical direction shown in
FIG. 3 represent the cosine measure of keywords included in the
detection windows set at adjacent positions on the time base. The
higher the value of the cosine measure is, the higher the
similarity value is.
[0054] As shown in FIG. 3, the positions for which the similarities
of the keywords extracted from the captions are low and below a
threshold value (for this embodiment) and the positions of the
changing points of topics confirmed by visual checks are almost
coincident, which shows that the method of this embodiment is
efficient and reliable. The processing of the information
processing apparatus 1 is described later with reference to a
flowchart.
[0055] FIG. 4 is a block diagram showing a hardware configuration
example of the information processing apparatus 1 of FIG. 1. The
hardware configuration shown in FIG. 4 may be part of a
microprocessor, chip, computer, personal digital assistant, a
mobile communication device, etc., as will be appreciated by one of
ordinary skill in the art. However, other possible implementations
of the information processing apparatus are possible, for example,
a specialized circuitry or software capable of running on a
microprocessor.
[0056] A CPU (Central Processing Unit) 11 executes various
processing in accordance with programs stored in a ROM (Read Only
Memory) 12 or a storage unit 18. Programs executed by the CPU 11,
data and so on are suitably stored in a RAM (Random Access Memory)
13. The CPU 11, the ROM 12, and the RAM 13 are mutually connected
by a bus 14.
[0057] An input and output interface 15 is also connected to the
CPU 11 through the bus 14. An input unit 16 receiving input of
contents and an output unit 17 outputting changing point
information are connected to the input and output interface 15.
[0058] The storage unit 18 connected to the input and output
interface 15 includes, for example, a hard disc, which stores
programs executed by the CPU 11 and various data. A communication
unit 19 communicates with external apparatuses through networks
such as Internet or local area networks.
[0059] A drive 20 connected to the input and output interface 15
drives removable media 21 such as a magnetic disc, an optical disc,
an electro-optical disc or a semiconductor memory, when they are
mounted on, and acquires programs and data stored therein. The
acquired program and data are forwarded to the storage unit 18 and
stored therein, if necessary.
[0060] FIG. 5 is a block diagram showing a function configuration
example of the information processing apparatus 1. At least a part
of function units shown in FIG. 5 is realized by a designated
computer executable program being executed by the CPU 11 of FIG.
4.
[0061] In the information processing apparatus 1, for example, a
text extraction unit 31, a keyword extraction unit 32, a detection
window setting unit 33, a similarity calculation unit 34 and a
changing point detection unit 35 are provided.
[0062] The text extraction unit 31 extracts text data (character
strings displayed as captions) from the supplied contents, and
outputs the extracted text data to the keyword extraction unit
32.
[0063] The keyword extraction unit 32 performs, for example,
morphological analysis with respect to text data supplied from the
text extraction unit 31, extracting morphemes of only a particular
part of speech such as nouns or verbs based on results of the
morphological analysis. The morphological analysis may include
dividing each sentence in corresponding words and associating each
word with a word class. This analysis may be further modified to
discard predefined word classes. For example, in one embodiment,
only the noun and verb classes are considered and all other parts
of the sentence are not considered.
[0064] The keyword extraction unit 32 regards the extracted string
of morphemes as the string of keywords (time series), outputting
the time series of keywords to the detection window setting unit 33
and the similarity calculation unit 34.
[0065] The detection window setting unit 33 sets the detection
windows (adjacent regions) surrounding keywords of prescribed
ranges which are adjacent on the time base, by shifting a keyword
one by one, and outputs information of the set detection windows
(information about the range of respective detection windows and
the like) to the similarity calculation unit 34.
[0066] The detection window may be set according to the prescribed
number so as to surround ten keywords as described above but also
set according to a prescribed number of sentences. For example, the
detection window may be set to surround keywords extracted from one
sentence displayed as a caption. Further, the detection window may
be set according to a predefined time interval to surround keywords
extracted from sentences displayed for a period of a prescribed
time (ten seconds for example) as a caption when the content is
reproduced.
[0067] A user may select how to set the detection windows, or the
user may select a granularity of the detection windows (in the
above example, the detection window may be set so as to surround 20
keywords or the detection window may be set so as to surround 50
keywords or any desired number of keywords). In addition, a
suitable type and granularity for the detection windows may be
found in advance according to categories of programs to be shown as
options, or the detection windows are automatically set based on
category information of the EPG (Electronic Program Guide) and the
like. Further, fineness in changes of the similarities to be
detected may be changed by changing the granularity of the
detection windows. Therefore, according to an embodiment of the
invention, the granularity is changed to a suitable setting in the
case of a later-described detection of a concentration degree of a
topic.
[0068] The similarity calculation unit 34 calculates the
similarities of keywords included in the detection windows set in
the time series. The keywords are supplied from the keyword
extraction unit 32, based on information supplied from the
detection window setting unit 33. The similarity calculation unit
34 outputs information of the calculated similarities to the
changing point detection unit 35.
[0069] In the similarity calculation unit 34, the similarities are
calculated not only based on the number of similar keywords as
described above but also based on, for example, the vector space
model or similar models. That is, the keywords included in
respective detection windows set at adjacent positions on the time
base are arranged on a high dimension vector space. The frequency
of appearance of respective unique keywords is represented in the
vector space for two adjacent detection windows to obtain one or
more vectors for each window. The vector difference between a
vector corresponding to a first window and a vector corresponding
to a second window may define the similarity value of the two
windows.
[0070] Alternatively, other correlations between the vectors
indicating the keywords included in one detection window and the
vectors indicating the keywords included in the other detection
window are calculated by using a scalar product, the cosine
measure, an Euclidian distance and the like, and the results
indicate the similarities of the topics of the two detection
windows.
[0071] The calculation methods of similarities are not limited to
the above examples. According to an embodiment of the present
invention, weighting is appropriately performed according to parts
of speech (word class) of keywords. For example, various weights
may be assigned to various parts of speech and the similarity value
of the topics of the detecting windows are calculated based on
these weights.
[0072] The changing point detection unit 35 detects a boundary
point between the detection windows as a changing point of topics
in the content. When a lower similarity value than a prescribed
threshold similarity value is calculated, based on information
supplied by the similarity calculation unit 34, the changing point
detection unit 35 outputs changing point information indicating the
position (time) of the detected changing point.
[0073] Processing of the information processing apparatus 1 which
detects the changing points of topics as described above is
explained with reference to a flowchart shown in FIG. 6.
[0074] In step S1, the text extraction unit 31 extracts text data
from the supplied content, and outputs the extracted text data to
the keyword extraction unit 32.
[0075] In step S2, the keyword extraction unit 32 performs the
morphological analysis with respect to the text data supplied from
the text extraction unit 31, and outputs the extracted time-series
of keywords, based on the result of the morphological analysis, to
the detection window setting unit 33 and the similarity calculation
unit 34.
[0076] In step S3, the detection window setting unit 33 sets the
detection windows surrounding keywords of prescribed ranges at
adjacent positions on the time base by shifting a keyword one by
one, in the case that keywords included in one detection window are
determined based on the number of keywords. Further, step S3
outputs information of the set detection windows to the similarity
calculation unit 34.
[0077] In the case that the keywords included in one detection
window are determined based on the sentence method, the detection
window setting unit 33 sets detection windows surrounding keywords
extracted from respective sentences by shifting a sentence
displayed as a caption one by one. Also, in the case that keywords
included in one detection window are determined based on the time
method, the detection window setting unit 33 sets detection windows
surrounding keywords extracted from sentences displayed as captions
at corresponding periods of time by shifting the time
sequentially.
[0078] In step S4, the similarity calculation unit 34 calculates
the similarities of the keywords included in the detection windows
supplied from the keyword extraction unit 32 based on information
supplied from the detection window setting unit 33, and outputs
information of the calculated similarities to the changing point
detection unit 35.
[0079] In step S5, the changing point detection unit 35 detects a
boundary point between the detection windows as a changing point of
the topics in the content. A lower similarity value than a
prescribed threshold similarity value is calculated based on
information supplied from the similarity calculation unit 34. The
changing point detection unit 35 outputs the changing point
information indicating the position of the detected changing
point.
[0080] According to the above processing, the information
processing apparatus 1 may calculate the similarities of respective
parts of the content by using text data, which is considered to
reflect the change of topics of the content. The information
processing apparatus 1 may also detect changing points of topics
based on the calculated similarities.
[0081] Next, a generation of attribute information based on the
changing point information generated as described above is
explained. The generated attribute information is added to the
content and used on occasions such as, for example, a reproduction
of the content.
[0082] FIG. 7 is a block diagram showing a function configuration
example of an information processing apparatus 101.
[0083] The information processing apparatus 101 may have the
hardware configuration of FIG. 4, similar to the above disclosed
information processing apparatus 1. An information processing unit
111 and an attribute information generation unit 112 may be
provided in the information processing apparatus 101 as shown in
FIG. 7 by a designated computer executable program being executed
by the CPU 11 of the information processing apparatus 101.
[0084] The information processing unit 111 input contents such as
television programs and movies as input, detects changing points of
topics appeared when the content is reproduced, and outputs
changing point information which is information indicating the
detected changing points to the attribute information generation
unit 112. That is, the information processing unit 111 may have the
same configuration as the configuration shown in FIG. 5, detecting
changing points in the manner described above.
[0085] The attribute information generation unit 112 generates
attribute information based on the changing points information
supplied from the information processing unit 111, and adds the
generated attribute information to the content inputted from
outside. The attribute information generation unit 112 sets, for
example, title chapters at positions detected as the changing
points of topics, and generates information indicating positions of
the chapters as attribute information.
[0086] In addition, the attribute information generation unit 112
selects keywords representing respective topics from the keywords
extracted by the information processing unit 111 based on text data
displayed as captions in sections of respective topics, and
generates information indicating the selected keywords (important
words) as attribute information. For example, a keyword may be
selected as an important word, which is most frequently included in
the keywords extracted from the text data displayed as a caption at
a section of each topic.
[0087] Attribute information generation processing of the
information processing apparatus 101 (the attribute information
generation unit 1 12) of FIG. 7 is explained with reference to a
flowchart shown in FIG. 8. The processing may be started, for
example, when the processing explained with reference to FIG. 6 is
performed by the information processing unit 111 and the changing
point information is supplied to the attribute information
generation unit 112.
[0088] In step S11, the attribute information generation unit 1 12
sets indicators of chapters at positions detected as the changing
points of the topics in the content based on the changing point
information supplied from the information processing unit 111, and
adds attribute information indicating positions of the chapters to
the content to be outputted.
[0089] Accordingly, an apparatus capable of reproducing contents
may automatically search reproducing positions based on the
positions of the chapters indicated by the attribute information,
or may display a window in which images of the positions where the
chapters are set are thumbnail-displayed.
[0090] Next, a generation processing of other attribute information
of the information processing apparatus 101 of FIG. 7 is explained
with reference to a flowchart shown in FIG. 9.
[0091] The processing may be also started, for example, when the
processing explained with reference to FIG. 6 is performed by the
information processing unit 111 and the changing point information
is supplied to the attribute information generation unit 112. From
the information processing unit 111, the time series of the
keywords are also outputted to the attribute information generation
unit 112.
[0092] In step S21, the attribute information generation unit 112
selects keywords representative of topics as important words in
respective sections of topics delimited by the changing points,
based on the changing point information supplied from the
information processing unit 111. The attribute information
generation unit 112 adds attribute information indicating important
words to the content to be outputted.
[0093] The important words selected as described above may be set
as titles of the detected respective sections, or used, for
example, when only sections in which character strings including
the important words are displayed as captions. According to an
embodiment of the invention, the attribute information generation
unit 112 may set the titles based on the important words or
generate information indicating the summarized reproduced
sections.
[0094] In the above embodiment, the similarities calculated based
on keywords included in the detection windows may be used for
detecting changing points of topics. However, the similarities may
be used for other purposes.
[0095] A similarity indicates that character strings including
similar keywords are sequentially displayed as captions. Therefore,
a part of a content having the high similarity value is considered
to be the part in which a topic is concentrated in a particular
subject. Accordingly, the similarity value can be also used as
being indicative of a concentration degree of a topic. In other
words, with regard to FIG. 10, one topic is illustrated having two
regions, one labeled "content of incident" and the other one
labeled "interviews . . . ." The high similarity value of the first
region comparative to the second region indicates that the first
region of the topic has the high concentration degree of newsworthy
information while the second region of the topic has secondary
information.
[0096] In another example, FIG. 11 shows one topic having multiple
regions Q1-Q5 and region Q5 having the highest number of keywords,
i.e., being the most attractive segment of the topic. The various
regions (segments) within a topic may be detected by using the
above discussed changing points method modified to determine the
minimum value of the cosine measure that are above (not below) the
threshold.
[0097] Thus, the minimum values that are below a predefined
threshold in this embodiment are used to determine the changing
points between topics, the minimum values above the predefined
threshold are used to determined the changing points between
segments with different concentration of keywords within the same
topic, and the maximum values determine those segments of the topic
that are the most attractive/newsworthy. Also, the most
attractive/newsworthy segments may be determined by selecting the
first "n" maximum values, with n being a positive integer
number.
[0098] By using the concentration degree of the topic discussed
above, for example, the concentration degrees of the topic in
respective topics delimited by the detected changing points are
compared, and a part/segment may be found in which the topic is
most concentrated in content (or a program).Thus, a part/segment
that includes a high concentration of keywords (part that is highly
representative of the topic) of the content may be determined.
[0099] When a granularity of the detection windows is set as
described above, a part in one topic in which the topic is
particularly concentrated may be found from transitions of the
concentration degrees of the topic within a topic section delimited
by the detected changing points.
[0100] As shown in FIG. 10, in a news program or the like,
generally, the concentration degree is high at an opening part when
a summary of the news is reported (a part of "content of an
incident" in FIG. 10), and the concentration degree is lower at a
part that may include interviews or comments (a part of "interviews
of neighbors and the like" in FIG. 10) continuously reported after
the summary of the news.
[0101] FIG. 11 is a graph showing an example of similarities
calculated when the detection of the changing points of the topics
are performed with respect to a quiz program.
[0102] As shown in FIG. 11, the quiz program can be delimited
according to a question (according to a topic) by the changing
points of topics detected as described above. In the example of
FIG. 11, the quiz program is delimited into respective sections Q1
to Q5 based on the similarities of the sections.
[0103] As can be seen from the transitions of the similarities, it
may be considered that the question Q5, in which the higher
concentration degree of the topic (similarity) is calculated, is
the most representative question in the quiz program. It is
preferable that when a digest play is performed, only the most
representative parts are reproduced.
[0104] According to an embodiment of the invention, the
concentration degree of the topic may also be a feature of the
content, and a most representative part may be detected by
combining features obtained based on video and audio. The feature
obtained based on video is, for example, that the number of scene
changes is large, and the feature obtained based on audio is, for
example, that sound volume is high.
[0105] In the above embodiment, the case for which the text data
from which the keywords are extracted is closed caption data as
explained above. However, the keywords may be extracted from the
text data obtained by recognizing character strings displayed on a
screen by open captions, and the extracted text data may be used
for calculation of similarities or detection of changing points in
the same way as for the case of the closed caption data.
[0106] The above processing may be executed by hardware, as well as
by software. When the series of processing is executed by software,
the software is installed from a program recording media in a
computer in which programs included in the software are
incorporated in dedicated hardware, or for example, in a
general-purpose computer which is capable of executing various
functions by installing various programs.
[0107] The program recording media stores programs to be installed
in the computer and includes, as shown in FIG. 4, the removable
media 21, which is a package media such as the magnetic disc
(including a flexible disc), the optical disc (including a CD-ROM
(Compact Disc-Read Only Memory), a DVD (Digital Versatile Disc)),
an electro-optical disc or a semiconductor memory, the ROM 12 in
which programs are stored temporarily or permanently, and a hard
disc forming the storage unit 18 and the like. The storage of the
programs on the program recording media is performed by using wired
or wireless communication media such as a local area network,
Internet, or digital satellite broadcasting through the
communication unit 19, which is an interface such as a router, and
a modem.
[0108] In the specification, the steps of describing programs
include not only the processing performed sequentially in time as
described above but also include processing not performed
sequentially in time but executed in parallel or individually.
[0109] According to an embodiment of the invention, the
similarities of the keywords, which may be a criteria for
positively detecting positions of changing points of topics in
contents, may be calculated.
[0110] It should be understood by those skilled in the art that
various modifications, combinations, sub-combinations and
alterations may occur depending on design requirements and other
factors insofar as they are within the scope of the appended claims
or the equivalents thereof.
* * * * *